chrome/browser/ash/chromebox_for_meetings/artemis

Last modified: 04/12/2024

Overview

This directory contains the source code for Artemis, a data-collection API maintained by the Chromebox-for-Meetings team that supports aggregating data from multiple sources and optionally uploading said data to a cloud logging platform for developer analysis. The API also supports adding "watchdogs" to any data source for action items that should be executed on specific events.

Components

• DataAggregator: responsible for fetching data from all sources and uploading data to cloud logging remote node.
• DataSource: represents a single source of data to be collected. Can be remote or local sources. Current supported (local) sources below:
  • LogSource: collects data from a single log file
  • CommandSource: collects output data from single command
• DataWatchDog: applied to a single DataSource. Executes a callback when chosen data source contains data that matches supplied filter.
• ERP node: remote endpoint that receives data using Chrome's Encrypted Reporting Pipeline protocol (see go/chromereportingdocs). This endpoint is managed by the Fleet team and will not be discussed here.

See mojom file for more detailed info: https://source.chromium.org/chromium/chromium/src/+/main:chromeos/services/chromebox_for_meetings/public/mojom/meet_devices_data_aggregator.mojom

Internal Notes

DataAggregator

• Manages multiple data sources
• Calls Fetch() on each DataSource on a periodic cadence. When the payload grows large enough, it will be pushed to the Fleet endpoint via ERP.
• If the upload succeeds, Flush() is called on corresponding DataSources to alert them that they can clear their internal buffers. If it fails, it will be re-attempted until success.

DataSource

• Collects data on its own (faster) cadence, separate from DataAggregator.
• Maintains an internal data buffer for new data. Calls to Fetch() will clear this data; the caller is responsible for ensuring that the data is transported properly.
• Internal buffer & separate cadence are used to support watchdogs. We want to poll for data much faster than the Fetch() cadence to ensure that (a) we trigger watchdog callbacks close to when the event occurs, and (b) we don't miss a pattern match altogether.
• If internal buffer is filled to its max limit, buffer fills will temporarily be halted to give the aggregator a chance to consume the data. This should only happen on repeated failures.

LocalDataSource

• An abstraction around a DataSource that serves data that can be obtained directly on the workstation
• Handles common operations, like internal buffer size capping, data redaction, watchdog validation, and upload preparation via structured data

CommandSource

• A type of LocalDataSource that collects output data from supplied command
• Internal buffer will only be appended to if the current output is not equal to the last-appended data, OR if a watchdog is added
• Supports both CHANGE and REGEX watchdogs

LogSource

• A type of LocalDataSource that collects output data from supplied log file
• Internal buffer will collect a chunk (N lines) from the file on each iteration
• Due to the nature of the data, LogSources only support REGEX watchdogs